Re: [PATCH bpf-next v6 4/7] bpftool: Implement minimize_btf() and relocations recording for BTFGen

[Andrii Nakryiko <andrii.nakryiko@gmail.com>]

On Wed, Feb 9, 2022 at 2:27 PM Mauricio Vásquez <mauricio@kinvolk.io> wrote:
>

It would be good to shorten the subject line, it's very long.

> minimize_btf() receives the path of a source and destination BTF files
> and a list of BPF objects. This function records the relocations for
> all objects and then generates the BTF file by calling btfgen_get_btf()
> (implemented in the following commit).
>
> btfgen_record_obj() loads the BTF and BTF.ext sections of the BPF
> objects and loops through all CO-RE relocations. It uses
> bpf_core_calc_relo_insn() from libbpf and passes the target spec to
> btfgen_record_reloc(), that calls one of the following functions
> depending on the relocation kind.
>
> btfgen_record_field_relo() uses the target specification to mark all the
> types that are involved in a field-based CO-RE relocation. In this case
> types resolved and marked recursively using btfgen_mark_type().
> Only the struct and union members (and their types) involved in the
> relocation are marked to optimize the size of the generated BTF file.
>
> btfgen_record_type_relo() marks the types involved in a type-based
> CO-RE relocation. In this case no members for the struct and union
> types are marked as libbpf doesn't use them while performing this kind
> of relocation. Pointed types are marked as they are used by libbpf in
> this case.
>
> btfgen_record_enumval_relo() marks the whole enum type for enum-based
> relocations.

It should be enough to leave only used enumerators, but I suppose it
doesn't take much space to record all. We can adjust that later, if
necessary.

>
> Signed-off-by: Mauricio Vásquez <mauricio@kinvolk.io>
> Signed-off-by: Rafael David Tinoco <rafael.tinoco@aquasec.com>
> Signed-off-by: Lorenzo Fontana <lorenzo.fontana@elastic.co>
> Signed-off-by: Leonardo Di Donato <leonardo.didonato@elastic.co>
> ---
>  tools/bpf/bpftool/Makefile |   8 +-
>  tools/bpf/bpftool/gen.c    | 452 ++++++++++++++++++++++++++++++++++++-
>  2 files changed, 454 insertions(+), 6 deletions(-)
>

Looks good, few nits and concerns, but it feels like it's really close
to being ready.

[...]

> +}
> +
> +struct btfgen_info {
> +       struct btf *src_btf;
> +       struct btf *marked_btf; // btf structure used to mark used types

C++ comment, please use /* */

> +};
> +
> +static size_t btfgen_hash_fn(const void *key, void *ctx)
> +{
> +       return (size_t)key;
> +}
> +
> +static bool btfgen_equal_fn(const void *k1, const void *k2, void *ctx)
> +{
> +       return k1 == k2;
> +}
> +
> +static void *uint_as_hash_key(int x)
> +{
> +       return (void *)(uintptr_t)x;
> +}
> +
> +static void *u32_as_hash_key(__u32 x)
> +{
> +       return (void *)(uintptr_t)x;
> +}
> +
> +static void btfgen_free_info(struct btfgen_info *info)
> +{
> +       if (!info)
> +               return;
> +
> +       btf__free(info->src_btf);
> +       btf__free(info->marked_btf);
> +
> +       free(info);
> +}
> +
> +static struct btfgen_info *
> +btfgen_new_info(const char *targ_btf_path)
> +{
> +       struct btfgen_info *info;
> +       int err;
> +
> +       info = calloc(1, sizeof(*info));
> +       if (!info)
> +               return NULL;
> +
> +       info->src_btf = btf__parse(targ_btf_path, NULL);
> +       if (!info->src_btf) {
> +               p_err("failed parsing '%s' BTF file: %s", targ_btf_path, strerror(errno));
> +               err = -errno;

save errno before p_err, it can clobber errno otherwise

> +               goto err_out;
> +       }
> +
> +       info->marked_btf = btf__parse(targ_btf_path, NULL);
> +       if (!info->marked_btf) {
> +               p_err("failed parsing '%s' BTF file: %s", targ_btf_path, strerror(errno));
> +               err = -errno;

same, always save errno first before any non-trivial function/macro call


> +               goto err_out;
> +       }
> +
> +       return info;
> +
> +err_out:
> +       btfgen_free_info(info);
> +       errno = -err;
> +       return NULL;
> +}
> +
> +#define MARKED UINT32_MAX
> +
> +static void btfgen_mark_member(struct btfgen_info *info, int type_id, int idx)
> +{
> +       const struct btf_type *t = btf__type_by_id(info->marked_btf, type_id);
> +       struct btf_member *m = btf_members(t) + idx;
> +
> +       m->name_off = MARKED;
> +}
> +
> +static int
> +btfgen_mark_type(struct btfgen_info *info, unsigned int id, bool follow_pointers)

id is type_id or could be some other id? It's best to be consistent in
naming to avoid second guessing like in this case.

> +{
> +       const struct btf_type *btf_type = btf__type_by_id(info->src_btf, id);
> +       struct btf_type *cloned_type;
> +       struct btf_param *param;
> +       struct btf_array *array;
> +       int err, i;

[...]

> +       /* tells if some other type needs to be handled */
> +       default:
> +               p_err("unsupported kind: %s (%d)", btf_kind_str(btf_type), id);
> +               return -EINVAL;
> +       }
> +
> +       return 0;
> +}
> +
> +static int btfgen_record_field_relo(struct btfgen_info *info, struct bpf_core_spec *targ_spec)
> +{
> +       struct btf *btf = (struct btf *) info->src_btf;

why the cast?

> +       const struct btf_type *btf_type;
> +       struct btf_member *btf_member;
> +       struct btf_array *array;
> +       unsigned int id = targ_spec->root_type_id;
> +       int idx, err;
> +
> +       /* mark root type */
> +       btf_type = btf__type_by_id(btf, id);
> +       err = btfgen_mark_type(info, id, false);
> +       if (err)
> +               return err;
> +
> +       /* mark types for complex types (arrays, unions, structures) */
> +       for (int i = 1; i < targ_spec->raw_len; i++) {
> +               /* skip typedefs and mods */
> +               while (btf_is_mod(btf_type) || btf_is_typedef(btf_type)) {
> +                       id = btf_type->type;
> +                       btf_type = btf__type_by_id(btf, id);
> +               }
> +
> +               switch (btf_kind(btf_type)) {
> +               case BTF_KIND_STRUCT:
> +               case BTF_KIND_UNION:
> +                       idx = targ_spec->raw_spec[i];
> +                       btf_member = btf_members(btf_type) + idx;
> +
> +                       /* mark member */
> +                       btfgen_mark_member(info, id, idx);
> +
> +                       /* mark member's type */
> +                       id = btf_member->type;
> +                       btf_type = btf__type_by_id(btf, id);
> +                       err = btfgen_mark_type(info, id, false);

why would it not follow the pointer? E.g., if I have a field defined as

struct blah ***my_field;

You at the very least would need either an empty struct blah or FWD
for struct blah, no?

> +                       if (err)
> +                               return err;
> +                       break;
> +               case BTF_KIND_ARRAY:
> +                       array = btf_array(btf_type);
> +                       id = array->type;
> +                       btf_type = btf__type_by_id(btf, id);
> +                       break;

[...]

> +err_out:
> +       bpf_core_free_cands(cands);
> +       errno = -err;
> +       return NULL;
> +}
> +
> +/* Record relocation information for a single BPF object*/

nit: missing space before */

> +static int btfgen_record_obj(struct btfgen_info *info, const char *obj_path)
> +{
> +       const struct btf_ext_info_sec *sec;
> +       const struct bpf_core_relo *relo;
> +       const struct btf_ext_info *seg;
> +       struct hashmap_entry *entry;
> +       struct hashmap *cand_cache = NULL;
> +       struct btf_ext *btf_ext = NULL;
> +       unsigned int relo_idx;
> +       struct btf *btf = NULL;
> +       size_t i;
> +       int err;
> +
> +       btf = btf__parse(obj_path, &btf_ext);
> +       if (!btf) {
> +               p_err("failed to parse BPF object '%s': %s", obj_path, strerror(errno));
> +               return -errno;
> +       }

check that btf_ext is not NULL?

> +
> +       if (btf_ext->core_relo_info.len == 0) {
> +               err = 0;
> +               goto out;
> +       }
> +

[...]